A model comparison of 2D Cartesian and 2D axisymmetric models for positive streamer discharges in air

TL;DR

This study compares 2D Cartesian and axisymmetric models of positive air streamers, revealing significant differences in inception voltage, streamer dimensions, and ionization, with implications for the accuracy of simplified simulations.

Contribution

It provides a detailed comparison highlighting the limitations of 2D Cartesian models for simulating positive streamer discharges in air.

Findings

2D Cartesian inception voltages are about twice those of axisymmetric models.

Streamers in 2D Cartesian are up to four times thinner and slower.

Branching cannot be qualitatively captured in 2D Cartesian geometry.

Abstract

Simulating streamer discharges in 3D can computationally be very expensive, which is why 2D Cartesian simulations are sometimes used instead, especially when dealing with complex geometries. Although 2D Cartesian simulations can only be used to obtain qualitative results, it is nevertheless interesting to understand how they differ from their 3D or axisymmetric counterparts. We therefore compare 2D Cartesian and axisymmetric simulations of positive streamers in air, using a drift-diffusion-reaction fluid model with the local field approximation. With the same electrode length and width, inception voltages are found to be about a factor two higher in the 2D Cartesian case. When compared at the same applied voltage, the 2D Cartesian streamers are up to four times thinner and slower, their maximal electric field is about 30% lower and their degree of ionization is about 65% lower, with the largest differences occurring at the start of the discharge. When we compare at a similar ratio of applied voltage over inception voltage, velocities become rather similar, and so do the streamer radii at later propagation times. However, the maximal electric field in the 2D Cartesian case is then about 20-30% lower, and the degree of ionization is about 40-50% lower. Finally, we show that streamer branching cannot qualitatively be modeled in a 2D Cartesian geometry.